Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20200917204514.GA2880159@google.com> <CAHk-=whXpv0KJvpL153dhUaRgSjzT8H4dD85hRw-fAwXvXnKAA@mail.gmail.com>
 <20200918162305.GB25599@embeddedor>
In-Reply-To: <20200918162305.GB25599@embeddedor>
From:   Linus Torvalds <torvalds@linux-foundation.org>
Date:   Fri, 18 Sep 2020 10:23:54 -0700
Message-ID: <CAHk-=wjH+OH08yjp=LpexkUnGp0Ogusk3WX0G7Q+Lh7Anfr21A@mail.gmail.com>
Subject: Re: [GIT PULL] percpu fix for v5.9-rc6
To:     "Gustavo A. R. Silva" <gustavoars@kernel.org>
Cc:     Dennis Zhou <dennis@kernel.org>, Tejun Heo <tj@kernel.org>,
        Christoph Lameter <cl@linux.com>,
        Linux-MM <linux-mm@kvack.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Fri, Sep 18, 2020 at 9:17 AM Gustavo A. R. Silva
<gustavoars@kernel.org> wrote:
>
> This bug could have been prevented by either adopting better
> coding practices or through the use[3] of the recent struct_size() helper.

Well, my unspoken point was that coding practices are just
theoretical. Coding practices don't help - actual *checking* of them
helps.

I realize that structures with flexible-array member are allowed to
use sizeof() in standard C, but if we want to make sure this doesn't
happen, we would need to have a stricter model than that. But a quick
google didn't find any flag to enable such a stricter mode.

I guess a sparse warning would work, but sparse already has too many
warnings and as a result most people don't care - even if they were to
run sparse in the first place.

Is there some gcc option that I didn't find to help find any questionable cases?

Because if we have a coding practice that you should use
'struct_size()', then we should also have a way to _verify_ that.

The whole - and really ONLY - point of using flexible arrays was that
it would protect against these things. And as things are now, it
simply doesn't. It's not an actual improvement over just using a
zero-sized array.

(Slightly related: copying a struct has the exact same issue. A
flexible array is no better than a zero-sized array, and generates the
same code and the same lack of any warnings, afaik).

               Linus